Just had a chance to have another look at the new algo.
It appears to work well. It's more consistent than before.
I'll need to fine tune the settings though.
Previously, I had dynamic tuning strength at under 40%. That might be too low with the new algo with the flat bright surfaces setting at the default 100% strength.
I also noticed that the FALL algo appears to be performing better than BT2390 with those settings for The Meg. I coloured the text red in the new algo heading for some scenes where the dynamic target was quite a bit higher when using BT2390 (ignore the final three text entries in red (338/374/374). I must have forgotten to change the text colour)
When I bumped the dynamic strength up to 50%, the discrepancy increased:
I was getting this for BT2390:
bt2390 100% flat/bright setting and 50% dynamic tuning strength
345 & 1375 dynamic tone map target
FALL 100% flat/bright setting and 50% dynamic tuning strength
302 & 761 dynamic tone map target
I think the algo could be improved further by separating two types of bright objects: bright flat areas, and bright flat areas with detail (although I'm not sure how this could be done based on histogram stats only. Maybe it might be easier in the live algo with pixel shader mapping? I've no idea tbh)
Here is a shot from Prometheus with FALL algo and 50% dynamic tuning strength
396 dynamic tone map target
there's still a little bit of a discrepancy (it's not too bad)
a later shot at 232 dynamic tone map target
Here is a shot from 2001: A Space Odyssey with FALL algo and 50% dynamic tuning strength
at 220 dynamic tone map target, which is too bright imo
contrast with live algo at 50%
at 551 dynamic tone map target